STP UplinkFast and BackboneFast

[Mitrixsen]

Hello, everyone.

When it comes to UplinkFast, my study resource says that if a device loses a root port, it can immediately have a blocking port be chosen as the next root port and transition into forwarding without going through the listening/learning states.

My question is, doesn't skipping the listening/learning states create the potential for a loop to occur? Because as far as I know, you transition through these states in order to learn and adapt to the STP topology. Or is the idea here that if you flip the root port, it only affects the local switch and not the downstream STP topology?.

Same goes for Backbonefast

However, in this case, if SW2 loses the G0/0 root port, SW3's G0/1 port will skip the Max Age timer but it will transition through the listening/learning states and not move directly to forwarding. Why? Why can Uplinkfast skip the 2 transitional states but Backbonefast cannot? Is it because in BackboneFast scenarios, the change also affects the downstream switches?

Thank you.

David

[Joseph W. Doherty]

What have been your study resources?

Have you read the Cisco Tech Notes on these technologies?

https://www.cisco.com/c/en/us/support/docs/lan-switching/spanning-tree-protocol/10575-51.html

https://www.cisco.com/c/en/us/support/docs/lan-switching/spanning-tree-protocol/12014-18.html

BTW, if you're studying STP, you might review the Cisco proprietary STP features (like the two above) compared to rapid STP.

Feel free to post follow up questions.

[Mitrixsen]

Thank you for the links. I do have a few questions.

At first when the STP topology is being built, there's no confirmation that everyone has agreed on what the STP topology should be, etc. Switches have to perform their port role selection, send out BPDUs, determine who the root bridge should be, and so on. So it it correct to say that STP plays it safe by implementing the learning and listening state? So you can adapt to the topology, pick exactly one path to the root bridge, and so on.

My understanding is that with UplinkFast, you already know what the STP topology is and if you do lose a root port, the only change that needs to happen is for your blocking port to move into forwarding which would happen regardless of whether there is a listening/learning state or not. In the end, you still have only one root port since the previous root port went down thus no loops should occur.

However, I am not quite sure why do you need to transition through these states with BackboneFast and not skip them.

Thank you.

[Joseph W. Doherty · Accepted solution]

"So it it correct to say that STP plays it safe by implementing the learning and listening state?"

Yes, although something like portfast bypasses such safety, which is why, generally it's only used on host edge ports.

Regarding your understanding of uplinkfast, that's correct.

As to backbone fast, the issue is indirect link outage.  Basically it's still working much like STP, but faster.  The key to the faster, though, is its own newer query process.

BTW, the backbonefast TechNote concludes that rapid STP already has this feature (because rapid is a more proactive approach rather than reactive).

[M02@rt37 · Accepted solution]

Hello @Mitrixsen  

When a switch with UplinkFast enabled loses its root port, it can immediately transition one of its blocked ports to the forwarding state without going through the listening and learning states. This is possible because UplinkFast is designed for access switches that have already determined potential backup root ports that are directly connected to the same segment or another switch's designated port. These backup ports are pre-configured to quickly take over as the root port, and the switch immediately starts forwarding frames while ensuring there are no loops by rapidly updating the bridge table with new path costs and notifying neighboring switches. This fast transition is localized to the switch itself and does not disrupt the overall STP topology, minimizing the risk of loops.

BackboneFast, on the other hand, addresses the scenario of an indirect link failure, such as a link failure in the backbone of a network. When Backbonefst is enabled, a switch detecting an inferior BPDU can infer that there has been a topology change and immediately start a process to find an alternative path to the root bridge. However, unlike Uplinkfast, BackboneFast does not immediately transition a port to the forwarding state. Instead, it bypasses the Max Age timer and quickly moves the port to the listening state, then proceeds through the learning state before finally reaching the forwarding state. The reason BackboneFast does not skip the listening and learning states is to ensure network stability and prevent loops, as the change involves multiple switches and potentially affects the entire STP topology. The listening state is crucial to remove any outdated MAC addresses, and the learning state allows the switch to build the correct MAC address table. By doing this, BackboneFast ensures that all switches in the topology correctly update their STP states and maintain a loop-free environment.